Application and Development of Artificial Intelligence in Forensic Science under the Background of Big Data

Ying Yuan1, Qian Dai2 1Department of Forensic Science, Guizhou Police College, Guiyang, China

2Department of Public Security, Guizhou Police College, Guiyang, China

Keywords: big data; artificial intelligence; criminal investigation

Abstract: With the advance of big data analysis technology, as well as the application of technologies including mode identification, in-depth learning, and computer vision technique, the artificial intelligence technology has welcomed ground-breaking progress. Under this background, the criminal investigation methods and emphases of forensic science have all changed, and various types of expert auxiliary system, crime prevention and alarm system, intelligent identity recognition system based on artificial intelligence are born. Although artificial intelligence still faces problems such as technical bottlenecks, irregular legal procedures, data collection authority when applied in forensic science area, it is believed to bring new breakthroughs to criminal investigation in forensic science in the future.

In nearly 20 years, with the rapid development of big data, machine learning, computer vision,

pattern recognition technology, forensic science research methods and research emphasis has changed. AImakebreakthrough in forensic science and criminal investigation as well as public security field. The concept of AI (artificial intelligence) was bornas early as in 1956, limited by computer hardware and computing algorithm storage level and other factors, AI development suffer chronic slow growth during this period. Until 2006, with the emergence of big data, cloud computing and other cutting-edgetechnologies, computing speed have improved rapidly. It also provided abundant data for AI, and assisted in training more intelligent training models. After that, machine learning algorithms, pattern recognition, machine learning achieved great success, making it possible to combine machine, human and network into a new group intelligence system.

1. Definition of artificial intelligence Artificial intelligence, the AI White Paper defines that AI is a theory, method, technology and

application system that use digital computers or digital control machine simulation, extend and expand people's intelligence, ambient intelligence, obtain data and use knowledge to analysisgets best results[2]. The concept of generalized AI can be divided into two forms: strong AI and weak artificial intelligence. Strong AI can simulate human intelligence and complete tasks that can only be completed by humans, such as learning new knowledge system alone, making choices in the face of uncertainty, and even having human consciousness and sense of existence [3]. Weak AI refers to the AI has not really have the independent consciousness. Along with the birth of big data, though the development of AI has made a phased breakthrough, what we concern in this paper is majority belongs to the category of the weak AI.

2. The necessity of application of AI in forensic science under the background of Big Data Traditional method of handling evidence is usually physical, chemical, or biological method to

develop, identify and link different kinds of marks and trajectory, behavior and material, combining with casing waiting, survey investigation way to narrowing the scope of investigation and to lock the criminal suspect in the forensic science field. At present, with the access of big data, the sources of data are more extensive. The sources of data are various, fast, multimodal, difficult to identify

2019 3rd International Conference on Artificial intelligence, Systems, and Computing Technology (AISCT 2019)

Copyright © (2019) Francis Academic Press, UK DOI: 10.25236/aisct.2019.087451

and low in value. The forms of data include image, number, voice and text. From the macroscopic view ,Big Data is a link that connect the Physical world, Information Space and the Human Society [1]. On the basis of Big Data access authority, AI can get a qualitative leap in the field of forensic science and make a breakthrough.

3. Application Forms of AI in Forensic Science under the Background of Big Data 3.1 Expert Auxiliary System

The main application field for criminal investigation intelligence is expert auxiliary system. Expert system is a kind of intelligent computer program system that simulatethe steps and methods according to the forensicscientistsin the working process ofcriminal investigation. At present our country have been trying to build expert auxiliary system used for the field of public security.

In China by the end of 2017, the Ministry of Public Security has access to 34 departments information including railway, civil aviation, justice, education, quality supervision, inspection,etc. In July, 2017, Shanghai established the first "Intelligent AuxiliaryCriminal Case Handling System" [4], also known as "206 system". A breakthrough from zero to one has been achievedwith the application of this system. This system carries out intelligent research and judgment of evidence intelligence through "entity relationship analysis technology", and conducts data analysis, feature extraction, machine learning training model and model evaluation. Linking material evidence and witness, public information for multidimensional data mining and intelligent analysisin crimescene. January 2018, Hangzhou Procuratorate have developed another intelligent auxiliary case system, transition from the traditional mode to digital case in trial to provide comprehensive and intelligent justice. 2017, in US, the federal bureau of investigation (FBI) and other law enforcement departments developed a variety of cloud platforms in combination with AI. The intelligent case handling auxiliary system is used in the field of forensic science, hoping that the intelligent system can deduce, make decisions and associate in this field like human beings, so as torealize real intelligent analysis. Experts in forensic science teams also gradually realize that AIis likely to improve law enforcement efficiency and reduce human resources and financial costs.

3.2 Crime prevention and control early-warning system The same type of crime has temporal and spatial regularityat some level. Big data technology

analyzes a series of case information such as robbery information, items involved in the case, records of former criminals and so on, summarizes the occurrence rules, means, methods, items involved in the case and the activity track of suspects of similar cases.Analyzing the relevant situation of multiple locations of the case within a certain probability range, and even analyzes the data of the victim, combines the information of criminal facts, can realize the purpose of early warning and anti-terrorism of the crime. Big Data crime prevention and control system enables us to pay attention to the data related to crime facts and not blindly explore the cause-and-effect relationship of crime. At present, some country, especially in developed area, have started to use such data to analyze the crime prevention and early warning system, improving the prevention of violent and terrorist crimes, financial crimes, cyber crimesand corruption crimes. In addition, the identification system for monitored populationdeveloped by the public security field is also used in crime prevention. The efficient face monitoring systemis adopted to help public security investigators to identify specific individuals, and to realize the real-time alarm functions of dangerous or criminalsidentification.

In 2017,IBMlaunchedi2 Coplink data processing software Mesa,Arizona Police Department. In 2019 they produceCoplink X that can search technology and cloud-based information capabilities to police department. It also incorporate the advanced analytical and data consolidation capabilities of the Coplink suite of products.Providing actual intelligence support law enforcement community and enhancing crime fighting efforts.Inresent years, the U.S. Department of Justice, the U.K. Financial Conduct Authority and other government law enforcement agencies have begun exploring advanced A.I. analysis tools with surveillance cameras, drones.Aimedatmaximize efficiency and the use of

452

evidence in crime detection, as well as promote impartiality in law enforcement and crime prevention. In addition, the Palantir technology company and the Los Angeles Police Department are develop a smart system cooperatively, to analyze the data integration andto predict management solutions. Intel corporation of the United Stateslaunched Saffron AML Advisor, which USES LiDAR algorithm to analyze and control financial crimes by accessing data sources ,such as large enterprises and the Internet. Advisor system can monitor financial crimes by simulating the thinking mode of human brain, and create highly transparent related information to solve the problems when facefinancial crimes, so as to help investigators and analysts understand and explore new fraud trends from the data of Banks or insurance companies.Visual analytics for sense-making in criminal intelligence analysis(VALCRI) is a software tool already being used by various police agencies. That is designed by a private American software company that specializes in big data analytics. The tool automatically searches the databases using dedicated search engines with one click.

3.3 Various Intelligent Personal Identification Systems Besides,AI application forms also include various intelligent personal identification systems,

such as automatic fingerprint recognition system(AFIS).AFIS in China started in the late 1970 s, the main company using fingerprint automatic identification are Peking University High Tech,Eastern Golden Finger, Beijing forensic science institution.etc. NewAIAFIS based on machine learning and pattern recognition algorithm, such astexture feature of wavelet framework presents adaptive scale and depth of the same characteristics of neural network. This algorithm make the details of the fingerprint characteristics won’t lost easily.Various research papers public in literally, such as: Convolution Neural Network (CNN) [5] Convolution Neural Network, Artificial Neural Network (ANN) Neural Network [6], Nonlinear Bback Propagation Neural Network (BPNN) reverse Neural Network [7], Linear Vector Quantization (LVQ) Linear Vector Quantization [8], Extreme Learning Machine extreme learning Machine (ELM) [9], etc. The applicationof large fingerprint identification system is relatively less. Qian rongxin et al. [10] from Ministry of public security use algorithm in his paper.The algorithm under the multi-node concurrent processing a large number of fingerprint data, and provides real-time query function and automatic feature extraction, extraction of fingerprint image information. The algorithm use the deep learning technology in the field of computer vision, image restoration and coding and adaptive wavelet frame technology, puts forward the new intelligent fingerprint image comparison system, can realize fully automated fingerprint feature extraction without human intervention.It can also identify incomplete fuzzy handprints, in the past those found at the scene of the crime of the imperfect fingerprints. If the number of fingerprint feature points in the fingerprint image does not meet the identification conditions, it cannot be identified, even if the fingerprint image lines are clear, the new fingerprint identification system can not be limited to the number of fingerprint detail feature points. The system can identify and match the incomplete finger and palm prints.

Intelligent gait retrieval system, The gait recognition is one ofthe new field in the feature recognition on biology. Human gait recognition is the process of identifying individuals by their walking manners. Traditional gait recognition technology use manual modeling and non-deep learning recognition algorithm, and it is difficult to achieve the continuity of feature recognition when suspect’s clothes changes in video, or covered by objects. The new AI gait recognition is based on such methods as deep learning discriminant method and generative method [11], gait energy map network[12], 3D convolution [13], key points of human posture [14], long and short time memory module and human joint heat map [15]. Gait feature recognition technology relies on monitoringpeople walking gait information and collectiongait features ,and compares the characteristics of the database, identify whether the characteristics belong to the same person [16]. The application of intelligent gait retrieval system relies on rich video image information. Based on the integration of video sharing platform, it provides rich video information for gait identification[17].In November 2018, Water Droplets Company, the Chinese academy of sciences, Milky Way cloud Company released the first global gait recognition retrieval intelligence , 1 hour of video can be finished in 10 minutes to retrieve, fastest will speed up the detection rate of many

453

kinds of cases.It is reported that at present the products have been used in public security system more than 1000 hours, the test to be involved in more than 20 case detection, total 2000G public security video retrieval, widely praise [18]. The advantages of non-contact and long-distance capture of gait recognition, it is believed that it will provide convenience to the public security field in the future.

Face recognition also has the advantages of non-contact, it’s disadvantage is the False Accept Rateis higher than fingerprint. In the first stage of face recognition, the distance between the eyes, nose and mouth is used as a parameter for recognition [19]. In this stage algorithm extracting features of five senses' location, distance parameter information is established to build a face recognition system. Second stage using the topology of face, matching face with key information. Matching face after the dimension reduction using linear discriminant analysis (LDA), thus the big difference between class and class differences within small linear subspace, shortcomings of this kind of method is unable to modeling of complex nonlinear model. In the third stage, Face recognition is based on video, near-infrared and other data sources to obtain Face data. Deep learning algorithm [20] is used. The most effective feature extraction method is LBP Face[21]. Along with the development of AI , the depth of the Belief Network DBN (Deep are Network [22], as well as the depth of the DCNN (Deep Convolution Neural Network) in the field of computer vision technology breakthrough, make the accuracy of face recognition, created the "brush face era". By the end of 2014, the popularity of the second generation of IDcards in China has enabled the identity information of 1.2 billion people to be provided with face digital photo data, and millions of video surveillance cameras connected with PinganCity Network .That have provided massive face data information in the field of public security [23]. Local public security departmentuse face recognitionto monitordangerous population,crime early warning, identity verification, electronic bayonet and criminal case investigation, etc.

Biometric identification systems including fingerprints, gait, face recognition,and iris recognition, micro expression recognition, ear shape recognition, speech recognition, vein recognition, etc. Iris area accounts for 65% of eyeball [24], contains pigment dots, pits, rich texture information, using machine learning methods, such as neural network and support vector machine (SVM) to extract iris feature matching and identity of finishing. Intelligent speech recognition technology includes speech recognition technology (ASR) and speech synthesis technology (TTS). Speech recognition in the field of public security is closer to voice print identification. Because of the character of person, the discrepancy of any two mens voice and the speaker can be determined with their voice print.If weadequately exert the function of voiceprint technology, it would increase the ratio of cracking criminal cases and find out new resource to facilitate the process of information detection[25]. The well-known Apple mobile phone Siri assistant, Google voice assistant Google Now, domestic Xiaomi Tech company AI intelligent speakers Xiao Ai Tongxue, Alibaba TmallTian Mao Jingling product of artificial intelligence, which bring significant convenience to our life. Micro-expression recognition is based on psychological research.Studies have shown that micro-expressions can accurate reflection of mood swings, while natural happy or sad expression can disguised[26]. Micro-expression exist a very short time, often only in exist for 0.2 seconds. Analysis of micro expression using the AImodel to change can achieve stability, the effect of early warning.Vein recognition technology has taken more attention in recent years with its unique advantages. Vein recognition is the use of infrared camera, the near infrared light-emitting tube and filter for vein image acquisition, image preprocessing, feature extraction and matching.At present is mainly used for identity recognition, security door special people in identifying features, feature matching method is different also. With face recognition, gait recognition, micro expression recognition mode depend on the amount of video data, all of the biometric identification technology depends on the advanced artificial intelligence, computer vision and image processing technology.

4. AI problems Under the background of big data highlighted the role of AIin forensic science field increasingly,

but many problems still exist.On the one hand, based on the existing conditions of science and

454

technology, human’s understanding of AIis limited. Understanding human intelligence itself is still in its primary stage. Humancan't completely create true artificial intelligence before make a deep understanding of human brain.Fully understanding human brain also needs a certain amount of. On the other hand, in terms of access to large amounts of data information involves a lot of problems. How to protect citizen’s privacy right? AIin the field of forensic science has five typical technology: Machine learning, computer vision, natural language processing, robotics, biometric identification technologyas well as natural language processing, pattern recognition, robot mechanical learning, statistical interference function and automatic programming, etc.This kind of technology make the processing of AImore efficient, and can process many kinds of fuzzy evidence, and obtain case clue information efficiently. These clues in the past need to spend huge manpower and financial resources and time to handle. With the accumulation of big data, the innovation of theoretical algorithm and the improvement of computer ability, artificial intelligence has been widely used in forensic science. Big data platform for AI provide huge amounts of data source, millions of neurons of human brain chip .Different kinds of deep learning processor provides the powerful computation ability for artificial intelligence. With the innovation of technology, the research object of criminal investigation is expanding. Technology increasingly advanced research method, which become more scientific and rigorous. We believe that the future development of AI will bring new breakthrough to forensic science field.

Acknowledgement Guizhou Provincial Department of Science and Technology basic research project: guizhou

science and technology foundation [2017]1063（Contract No） Guizhou Provincial Department of Education college humanities and social sciences research

youth project :2018qn03（Contract No） Guizhou Provincial Department of Education youth science and technology talent growth project:

guizhou education KY [2018]289（Contract No）

References [1] Chengxueqi, jinxiaolong, wang yuanzhuo. Overview of big data system and analysis technology [J], journal of software, 2014:9 (25) : 1890-1908. [2] China institute of electronic technology standardization, tsinghua university, Peking University, etc. White paper on AI[M], Beijing: China institute of electronic technology standardization, 2018:5-6. [3] Maximilian Nominacher,BertrandPeletier. AIPolicies[M],US:Wiley Press,2017:64-65. [4] The country's first "criminal case intelligent auxiliary case system" was born in Shanghai, People's Daily online [EB/OL]. [2017-07-10]. Http://sh.people.com.cn/n2/2017/0710/c134768-30446516.html [5] Peralta, D., Triguero, I., Garcia, S., Saeys, Y., Benitez, j. M., & Herrera, F. (2018). On the use of convolutional neural networks for robust classification of multiple fingerprint captures. International Journal of Intelligent Systems, 33 (1), 213-230. The doi: 10.1002 / int. 21948 [6] Adjimi, A., hacine-gharbi, A., Ravier, P,el at. Extraction and Selection of Binarised Statistical Image Features for Fingerprint Recognition[J]. International Journal of Biometrics,2017,9(1):67 -- 80. [7] Yang, j. C., Park, d. s. Fingerprint Verification Based on Invariant Moment Features and Nonlinear BPNN[J].International Journal of Control, 2008,6(6):800-808. [8] Kumar, Fingerprint Matching Using quality Orientation Local Binary Pattern Descriptor and Machine Learning Techniques[J]. International Journal of Computer Vision and Image Processing

455

2017, 7 (4): 51-67. [9] Fingerprint Matching Based on Extreme Learning Machine[J]. Neural Computing & Applications, 2013,22(34):435 -- 445. [10] Qian rongxin. Application of distributed fingerprint comparison system in the public security industry [J]. Information and computer, 2017(8):131-132.

456